In the last decade, advances in functional magnetic resonance imaging (fMRI) have shifted the focus of cognitive neuroscience research toward understanding how memory traces are represented in the human brain. In their thoughtful synthesis of recent neurocognitive models, Bastin et al. illustrate the important role that attribution plays in recollection and familiarity, and incorporate multiple factors to put forth a holistic account of episodic memory. However, while the attribution system described in their integrative memory model is demonstrably essential for recollection and familiarity, it is secondary to the underlying memory representations.

Although judgments of recollection and familiarity are the measurable outputs in episodic memory tests, objective memory traces may sometimes be biased by the attribution system. For example, in a simultaneous fMRI and eye-tracking study, Hannula and Ranganath (Reference Hannula and Ranganath2009) reported that hippocampal activation predicts eye gaze to the correct item during a relational memory test, even when limited to incorrect explicit memory responses (i.e., implicit memory). Moreover, functional connectivity between hippocampus and lateral prefrontal cortex was greater for correct than incorrect responses. Together, these results tie episodic memory traces to implicit behavior (i.e., eye gaze), unbiased by attribution, as well as to explicit behavior (i.e., memory response), biased by attribution signals from prefrontal cortex.

As Bastin et al. discuss, the entity representation and relational representation core systems are recruited in the service of cognitive tasks beyond episodic memory (Graham et al. Reference Graham, Barense and Lee2010; Yonelinas Reference Yonelinas2013). Notably, there is evidence of shared neural substrates (Wang & Giovanello Reference Wang and Giovanello2016) and cognitive mechanisms (Wang & Yonelinas Reference Wang and Yonelinas2012) between explicit forms of memory such as recollection and familiarity, and implicit forms of memory such as priming. In other words, entities and relations may not need to be explicitly retrieved in order to involve recollection- and familiarity-based mechanisms. Consistent with the importance of these core systems for mnemonic representations, medial temporal lobe activation is related to objective oldness, rather than subjective or perceived oldness (Daselaar et al. Reference Daselaar, Fleck, Prince and Cabeza2006).

Therefore, in studying memory-impaired populations, it is essential to separate an inability to form mnemonic representations from attributional biases that affect the use of these representations. For example, memory-impaired patients with intact attribution systems may interpret processing fluency differently than healthy controls (Ozubko & Yonelinas Reference Ozubko and Yonelinas2014). The entity and relational systems are the backbones by which memories are encoded, and their retrieval can be most accurately assessed by cutting out the middleman – through implicit measures that do not rely on attributional systems. Moreover, separating attributional biases from memory deficits will also help to resolve inconsistent findings in the literature with regard to clinical populations. This perspective largely agrees with how Bastin et al. have described the integrative memory model. However, separating attributional biases is an important point of emphasis with both theoretical implications for how recollection- and familiarity-based memory is conceptualized, as well as practical implications for how residual memory function can be best harnessed in clinical populations.